Large-scale industrial energy systems optimization under uncertainty: A data-driven robust optimization approach

Abstract

In the large-scale industries, optimization of multi-type energy systems to minimize the total energy cost is of great importance and has received worldwide attentions. In the real industrial plants, the deterministic optimization may encounter difficulties because of various uncertainties. In this paper, the deterministic and robust optimization frameworks are proposed for energy systems optimization under uncertainty. A hybrid modeling method is applied to develop building block models based on the mechanism and process historical data. The deterministic optimization model can be further formulated as a mixed-integer non-linear programming problem. Considering enthalpy uncertainties, a generalized intersection kernel support vector clustering is employed to construct the uncertainty set. By introducing the derived uncertainty set in the deterministic optimization model, a robust optimization model is presented. A case study on the energy system of a real ethylene plant is carried out to illustrate the performance of the proposed approach and the effect of regularization parameter κ on the optimization results is studied. The results show that the optimized energy costs are 15148.84 kg/h and 16209.81 kg/h in deterministic and robust optimization methods. Despite higher energy consumption in robust optimization, the proposed method yields a trade-off between energy cost and robustness. The conservatism of the solution can be adjusted by the regularization parameter, and in this system κ=0.02 is recommended.